Theoretical Calculations to Assist Experimental Crystal Form - PowerPoint PPT Presentation

Theoretical Calculations to Assist Experimental Crystal Form Screening Jacco van de Streek University of Copenhagen Department of Pharmaceutics and Analytical Chemistry

Crystal Structure Prediction: Basics Free energy (p,T) energy structure ranking generation

Why Force Fields Do not Work Typical accuracy Required accuracy (0.1 – 0.01 kcal/mol) energy

Dispersion-corrected DFT (DFT-D) Force fields... Predicted RMS = 0.497 Å Experiment

Dispersion-corrected DFT (DFT-D) Pure DFT... RMS = 0.833 Å Force fields... Predicted RMS = 0.497 Å Experiment

Dispersion-corrected DFT (DFT-D) Dispersion-corrected DFT... RMS = 0.084 Å Pure DFT... RMS = 0.833 Å Force fields... Predicted RMS = 0.497 Å Experiment

Dispersion-corrected DFT (DFT-D) E lattice = E DFT + E vdW VASP • PAW potentials • Pair potentials • Plane-wave basis set • Element dependent • GGA – PW91 / GGA – PBE • Hybridisation dependent • - C 6  r -6 • 520 eV enery cut-off • 0.07 Å -1 k-point spacing “0 K” calculations, no free energies M. A. Neumann & M.-A. Perrin (2005) J. Phys. Chem. B 109 , 15531-15541 G. Kresse & J. Hafner (1993) Phys. Rev. B 47 , 558-561

Crystal Structure Generation Parallel tempering Monte-Carlo algorithm Tailor-made force field: accurate force field, fitted to artificial DFT-D reference data for each compound One and two independent molecules in all 230 space groups M. A. Neumann (2008) J. Chem. Phys. B 112 , 9810-9829

Examples: Blind Tests Crystal Structure Prediction Blind Tests: Blind Tests in 1999, 2001, 2004, 2007 and 2010 Good compounds for validation G. M. Day et al . (2009) Acta Cryst. B 65 , 107-125

Previous Blind Test Compounds :

Ranking Results ( NOT Full Studies) 1 & 2 2 2 1 1 1 1 & 2 1 1 4 1 1 & 2 1 1 A. Asmadi, M. A. Neumann, J. Kendrick, P. Girard, M.-A. Perrin & F. J. J. Leusen (2009) J. Phys. Chem. B 113 , 16303-16313

Previous Blind Test Compounds Conclusion Dispersion-corrected DFT appears to work well for energy-ranking crystal structures (80% success rate)

Efavirenz Number of patents: 5 Number of claimed forms: 24 Number of crystal structures known: 0

Efavirenz Experimental polymorph screen by Dr Eva Dova (Avantium) Scan of patent literature by Dr Menno Deij (Avantium) The 24 forms boil down to eight distinct forms: A, B, C, D, E, F, H1, N

Efavirenz ( Z '=1-2) A B C D E F H1 N Density Energy

Efavirenz ( Z '=1-2) A B C D C E F H1 F N B B B Density B N H1 Energy

Efavirenz ( Z '=1-2) A Some patents: C most stable (@RT) B Some patents: F most stable (@RT) C D C E F H1 F N B B B Density B N H1 Energy

Efavirenz Form B is disordered, four orientations of the cyclopropane group are found in the search (ranks 20, 27, 40, 57) Disorder means that our 0 K energies are not reliable S. Cuffini, R. E. Howie, E. R. T. Tiekink, J. L. Wardell & S. M. S. V. Wardell (2009) Acta Cryst. E 65 , o3170-o3171

- Form N is Z '=2 - One axial -CF 3 , one equatorial -CF 3 - Requires fully flexible search with Z '=2

Efavirenz Rietveld refinement with TOPAS for form N ( Z '=2) We can solve structures from poor quality laboratory powder patterns, scanned from a patent: low resolution, preferred orientation.

What about Forms A, D & E? A B C D C E F H1 F N B B B Density B N H1 Energy

Efavirenz A Z '=3 / Z '=6 [1] B Found, Disordered [2] C Found D Solvate (from TGA) E ? F Found [3] H1 Found N Found [1] S. Mahapatra, T. S. Thakur, S. Joseph, S. Varughese & G. R. Desiraju (2010) Cryst. Growth Des . 10 , 3191-3202 [2] S. Cuffini, R. E. Howie, E. R. T. Tiekink, J. L. Wardell & S. M. S. V. Wardell (2009) Acta Cryst. E 65 , o3170-o3171 [3] K. Ravikumar & B. Sridhar (2009) Mol. Cryst. Liq. Cryst . 515 , 190-198

Efavirenz Conclusions - Real life very complicated: - Z '>2 - Disorder: entropy contribution - Solvates - Form E ambiguous - Crystal structures from very poor quality powder patterns: use crystal-structure prediction - No more stable form found: it is unlikely that one turns up in the future

Current complexity limit F HO O O F H N S F O F Bicalutamide N Blind Test XX - Six months on a 64 CPU quad-core Xeon or Opteron cluster - Quasicomplete screen for Z '=1 in 230 space groups - 50 – 90 % complete screen for Z '=2 in 230 space groups

Acknowledgements Marcus Neumann - Avant-garde Materials Simulation Eva Dova - Avantium Menno Deij - Avantium